BrainBot Begins

[JonHylands]

Hi guys,

While waiting for the laser-cut surface mount soldering stencils for my IMU board (they shipped yesterday from Pololu), I did some work on building the new chest compartment for the robot. BrainBot is the Bioloid humanoid robot I am building for the Brain Engineering Lab at Dartmouth College in New Hampshire.

They have provided me with a Sherline CNC milling machine, which makes cutting complex shapes and patterns possible, pretty much directly from CAD.

Anyways, I updated the wiki page for BrainBot with some pictures:

http://www.bioloid.info/tiki/tiki-index.php?page=BrainEngineering+BrainBot

As I mention on that page, we found that the wifi module we bought just doesn't seem to work, and rather than waste a lot of time going through tech support, we're taking advantage of the timing and so we picked up a new gumstix verdex board, which has (among other things) host USB.

- Jon

[limor]

I'm amazed at the amount of productive work you manage to churn out the past few months.
Please document your experiences with the new PXA270 Gumstix.
Didn't the 270 have floating point co-processor ("mobile MMX" it was called a couple of years ago when it came out) ?

I've managed to kill my 2 Robostix boards and just sent them for RMA. So i'll attempt to communicate the Gumstix with the Bioloid using the Tweener board.

[JonHylands]

Heh, I wish I had a solid 40 hours a week to work on this stuff. This is just my spare time...

I will be fully documenting and open-sourcing everything I do with respect to BrainBot (part of the agreement with the Brain Engineering Lab).

I don't know about the floating point - for now we are just going to use the gumstix as a fast wifi <-> Bioloid bus pipeline. All the real processing will take place on the PC. I may end up writing the equivalent of an extended command (SYNC_READ), to allow reading values from a bunch of AX-12 sensors at once without needing to go back and forth across wifi for each one.

I would be surprised if you could talk to the Bioloid bus through the tweener, unless you're reducing the baud rate and using a buffer/inverter combo. I believe the hardware UART maxes out at 921 kbps.

- Jon

[JonHylands]

Well, its time for an update - after much machining, the chest is pretty much done...





The robot weighs in at a hefty 3.0 kg (the two batteries and the camera together weight almost a kilo).

We've determined that the FT232 drivers are installed in the gumstix build, but they aren't enabled, so my brother is rebuilding the system with that fixed for me, and hopefully I'll have gumstix control of this thing later on this week.

I've ordered some new grippers, and once I get the PCB sorted out for those, and figure out how to pull the second USB port from the 24-pin flex connector on the gumstix, this thing will be almost ready.

I'm going down to New Hampshire at the end of May to give a demo of BrainBot to the Brain Engineering Lab (the people funding the project), so its all coming down to the wire, but I think it will be ready.

Once I get everything working and have it walking, I'll post some video on my site, and let everyone know...

- Jon

[JonHylands]

So, last night I reached a major milestone with BrainBot - I managed to move a servo, untethered. Doesn't sound like much, but there's a lot of technology running to make that happen.

BrainBot, in the video below, is running off batteries (two separate battery packs - one for the electronics/gumstix and one for the bus). I have an ssh window on my laptop connected via wifi to the gumstix, which has a usb wifi dongle attached to one of the two USB host ports. I have an FT232 chip also onboard, connected to the second USB host port on the gumstix, allowing the gumstix to talk to the Bioloid bus at 1.0 Mbps directly, as a virtual serial port.

http://www.bioloid.info/BrainBot-Untethered.wmv (1.3 MB)

I'm running Squeak Smalltalk on the gumstix, which sends a command to the head tilt servo to change the position of the servo a couple times.

So, what I have now is a high-bandwidth, low-latency connection to the Bioloid bus from a PC. I have a colleague who has written a "pipeline" application (also in Squeak) that acts as a message forwarder between the Bioloid bus and a TCP/IP socket. This message forwarder also has implemented (on the gumstix) a SYNC_READ command, which allows me to read specific values from each actuator and sensor on the bus, in one message/response.

Once I get this pipeline software integrated, I will be able to use my old GUIs and motion control systems (which I wrote to make MicroRaptor walk) to talk to BrainBot.

- Jon